· Venture Capital Tracker · investment-strategies · 2 min read
AIR’s $50M Seed: Sequoia Then Greenoaks Bet on a Firewall for AI Agents
AIR emerged from stealth with $50M across two seeds — Sequoia-led $10M then Greenoaks-led $40M — to discover agents, continuously re-vet skills/MCP add-ons, and block unsafe connections.
VCT data record
Funding event facts
Source-backed financing and transaction details. Unknown terms remain undisclosed rather than estimated.
AIR raises $50M across two seed rounds (Sequoia, Greenoaks)
Tel Aviv/NY-associated AI agent security startup AIR emerged with $50M — Sequoia-led $10M seed then Greenoaks-led $40M — to build an inline firewall for agent skills and MCP supply chains.
- Event type
- Funding Round
- Event date
- Sep 1, 2026
- Stage / label
- Seed
- Amount
- $50M
- Confidence
- Company Disclosed
Company / target: AIR
Lead: Sequoia Capital , Greenoaks
Participants: Swish Ventures , Netz Capital
Sources: techcrunch.com
Sequoia led a $10 million seed and Greenoaks led a follow-on $40 million seed — $50 million total — as AIR emerged from stealth on September 1, 2026 (TechCrunch). Swish, Netz, and angels including Cognition’s Zach Frankel, Wiz’s Yinon Costica, Eon’s Ofir Ehrlich, Clay’s Varun Anand, and Anne Neuberger participated.
Spine: agent adoption is creating a new software supply chain (skills, plugins, MCP servers). AIR bets enterprises need an inline firewall that re-verifies those components continuously — not a static scan.
Key facts
| Field | Detail |
|---|---|
| Company | AIR / AIR Security (founders Yair Saban, Niv Hoffman — Unit 8200) |
| Round | Two seeds: $10M (Sequoia) + $40M (Greenoaks) |
| Date | Announced September 1, 2026 |
| Headcount | ~40 (TechCrunch) |
| Traction (company) | >20 customers; ~25% large enterprises; FS and pharma strongest |
| Claim | Filters out ~27% of add-ons/skills found online |
Who uses the product — and for what job
Users: security and IT teams in enterprises rolling out agent fleets — especially regulated industries that cannot treat MCP add-ons like unsigned drivers.
Job: discover agents (including shadow AI), intercept skill/tool loads, compare against AIR’s continuously updated whitelist, and block unsafe connections.
Why now
- Agents increasingly act like OSes with installable “drivers,” without the signing culture of modern kernels (Saban’s analogy to TechCrunch).
- Attackers can poison content and dependencies agents fetch rather than hacking the model directly.
- Category capital is already large elsewhere (Zenity’s $125M Series C in August; Noma’s prior $100M Series B) — AIR is racing for the continuous vetting wedge.
Why Sequoia — portfolio fit
Sequoia partner Bogomil Balkansky: this is an infrastructure re-verification problem before it is a scanner problem. Same-day Sequoia also spun out Empirik for change-risk prediction — a coherent “make autonomous software safe to run” theme.
| Dimension | Fit |
|---|---|
| Stage | Outsized seed to buy category time |
| Thesis | Control plane for agent supply chains |
| Risk | Platform bundling by labs/cloud; crowded agent-security field |
Greenoaks, Swish, and Netz are covered without inventing /fund/ pages.
What remains undisclosed
- Valuation / ownership after the stacked seeds
- ARR or ACV
- How the whitelist is audited independently
- Exact HQ footprint beyond Israel founding story / NY newswire dateline
Competitive map
Noma, Zenity, Astrix, Operant and others also sell agent/MCP governance. AIR’s stated moat is continuous ecosystem vetting, not endpoint visibility alone.
Takeaway
$50M at seed finances a standards race. The unresolved question: does “agent firewall” stay a durable category, or get absorbed into identity, SSE, and cloud security suites once MCP patterns stabilize?
Sources
Follow Venture Capital Tracker in Google
Add VCT as a preferred source to make our venture-capital coverage easier to find in Google Search.
Editorial note: AI tools assisted with research, structure, or drafting. Venture Capital Tracker retains human editorial responsibility for factual accuracy, relevance, and source quality before publication.
